Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Search US and EU case law, court opinions, dockets, and regulations via CourtListener, Harvard Case Law, and EUR-Lex. Use when: (1) finding US court opinions by keyword or citation, (2) searching federal and state dockets, (3) querying EU legislation and case law, (4) looking up cases by date range or jurisdiction. NOT for: current legislation text (use congress.gov), legal advice (never provide), patent search (use USPTO APIs).
.claude/skills/thomasmoreai-legal-search/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-08 | ✗→✓ | ▲ Improved | 19% | 0% |
| case-11 | ✗→✓ | ▲ Improved | 341% | 0% |
| case-15 | ✗→✓ | ▲ Improved | 80% | 0% |
| case-10 | ✓→✓ | = Same ✓ | 17% | 0% |
| case-02 | ✓→✓ | = Same ✓ | 48% | 0% |
Search US case law via CourtListener and Harvard Case Law Access Project, and EU law via EUR-Lex. Covers court opinions, dockets, and legislative documents.
Free, open API for US federal and state court opinions and dockets. Base URL: https://www.courtlistener.com/api/rest/v4/
bashcurl -s "https://www.courtlistener.com/api/rest/v4/search/?q=qualified+immunity&type=o&order_by=score+desc" \ | python3 -c " import sys, json data = json.load(sys.stdin) for r in data.get('results', [])[:10]: name = r.get('caseName', 'N/A') court = r.get('court', 'N/A') date = r.get('dateFiled', 'N/A') cite = r.get('citation', [r.get('sibling_ids', 'N/A')]) print(f'[{date}] {name}') print(f' Court: {court}') print() "
bashcurl -s "https://www.courtlistener.com/api/rest/v4/search/?q=antitrust+merger&type=r&order_by=score+desc"
bashcurl -s "https://www.courtlistener.com/api/rest/v4/search/?q=fourth+amendment+digital+privacy&type=o&filed_after=2020-01-01&filed_before=2024-12-31&order_by=dateFiled+desc"
bash# Supreme Court opinions curl -s "https://www.courtlistener.com/api/rest/v4/search/?q=free+speech&type=o&court=scotus&order_by=dateFiled+desc" # Specific circuit court curl -s "https://www.courtlistener.com/api/rest/v4/search/?q=patent+eligibility&type=o&court=cafc&order_by=dateFiled+desc"
scotus (SCOTUS), ca1-ca11 (Circuit Courts), cafc (Federal Circuit), cadc (DC Circuit), nyd (SDNY), cand (N.D. Cal).
Historical US case law (1658-2020). Free API with registration. Base URL: https://api.case.law/v1/
bashcurl -s "https://api.case.law/v1/cases/?search=miranda+rights&decision_date_min=2000-01-01&page_size=10" \ | python3 -c " import sys, json data = json.load(sys.stdin) for c in data.get('results', []): name = c.get('name_abbreviation', 'N/A') date = c.get('decision_date', 'N/A') court = c.get('court', {}).get('name', 'N/A') cite = c.get('citations', [{}])[0].get('cite', 'N/A') if c.get('citations') else 'N/A' print(f'[{date}] {name}') print(f' Citation: {cite} Court: {court}\n') "
bashcurl -s "https://api.case.law/v1/cases/?cite=410+U.S.+113"
bashcurl -s "https://api.case.law/v1/cases/?search=eminent+domain&jurisdiction=cal&decision_date_min=2010-01-01&page_size=10"
Access EU legislation, case law, and treaties via the EUR-Lex SPARQL endpoint.
bashcurl -s -G "https://eur-lex.europa.eu/EURLexWebService" \ --data-urlencode "query=SELECT ?doc ?title WHERE { ?doc a <http://publications.europa.eu/ontology/cdm#regulation> ; <http://publications.europa.eu/ontology/cdm#resource_legal_title> ?title . FILTER(CONTAINS(LCASE(?title), 'artificial intelligence')) } LIMIT 10" \ --data-urlencode "format=application/sparql-results+json"
bashcurl -s "https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:32016R0679" -o gdpr.html
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-10 | pass→pass | 8,983 | 4,766 | -47% | 1 | 1 | 0% | 2,037 | 2,385 | +17% | 0 | 0 | — |
case-01 | fail→fail | 18,862 | 8,367 | -56% | 1 | 1 | 0% | 4,707 | 1,851 | -61% | 0 | 0 | — |
case-02 | pass→pass | 12,496 | 11,139 | -11% | 1 | 1 | 0% | 2,634 | 3,900 | +48% | 0 | 0 | — |
case-03 | pass→pass | 9,297 | 10,557 | +14% | 1 | 1 | 0% | 2,107 | 2,909 | +38% | 0 | 0 | — |
case-04 | pass→pass | 7,072 | 6,213 | -12% | 1 | 1 | 0% | 1,503 | 2,662 | +77% | 0 | 0 | — |
case-05 | pass→pass | 7,542 | 4,336 | -43% | 1 | 1 | 0% | 1,670 | 2,369 | +42% | 0 | 0 | — |
case-06 | fail→fail | 13,242 | 8,382 | -37% | 1 | 1 | 0% | 2,623 | 2,423 | -8% | 0 | 0 | — |
case-07 | pass→pass | 7,091 | 3,637 | -49% | 1 | 1 | 0% | 1,492 | 2,151 | +44% | 0 | 0 | — |
case-08 | fail→pass | 7,969 | 4,604 | -42% | 1 | 1 | 0% | 1,880 | 2,246 | +19% | 0 | 0 | — |
case-09 | pass→pass | 8,975 | 4,545 | -49% | 1 | 1 | 0% | 2,095 | 2,407 | +15% | 0 | 0 | — |
case-11 | fail→pass | 2,415 | 3,145 | +30% | 1 | 1 | 0% | 439 | 1,937 | +341% | 0 | 0 | — |
case-12 | pass→pass | 3,112 | 1,510 | -51% | 1 | 1 | 0% | 632 | 1,624 | +157% | 0 | 0 | — |
case-13 | pass→pass | 8,068 | 5,560 | -31% | 1 | 1 | 0% | 2,131 | 2,365 | +11% | 0 | 0 | — |
case-14 | pass→pass | 7,634 | 3,917 | -49% | 1 | 1 | 0% | 1,740 | 2,236 | +29% | 0 | 0 | — |
case-15 | fail→pass | 4,609 | 1,980 | -57% | 1 | 1 | 0% | 940 | 1,690 | +80% | 0 | 0 | — |
case-16 | pass→pass | 8,412 | 3,405 | -60% | 1 | 1 | 0% | 1,815 | 2,095 | +15% | 0 | 0 | — |
case-17 | pass→pass | 12,403 | 12,133 | -2% | 1 | 1 | 0% | 2,092 | 3,442 | +65% | 0 | 0 | — |
case-18 | pass→pass | 7,781 | 4,834 | -38% | 1 | 1 | 0% | 1,722 | 2,438 | +42% | 0 | 0 | — |
case-19 | pass→pass | 4,888 | 2,259 | -54% | 1 | 1 | 0% | 995 | 1,794 | +80% | 0 | 0 | — |
case-20 | pass→pass | 13,198 | 9,058 | -31% | 1 | 1 | 0% | 1,988 | 2,977 | +50% | 0 | 0 | — |
case-21 | fail→fail | 27,916 | 24,415 | -13% | 1 | 1 | 0% | 5,134 | 6,203 | +21% | 0 | 0 | — |
case-22 | pass→pass | 5,302 | 4,925 | -7% | 1 | 1 | 0% | 935 | 2,166 | +132% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 21 counted toward the lift figure. The other 1 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +14 percentage points is the difference between those two pass rates over the 21 comparable cases.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.